Data as an Asset: The Friedkin Group’s Road to AI Readiness

Chad Watt:

Welcome to the Infosys Knowledge Institute Podcast where business leaders share what they've learned on their technology journey. I'm Chad Watt, Infosys Knowledge Institute researcher and writer.

Today I'm speaking with Ken Elliott, Chief Data and Analytics Officer at the Friedkin Group.

The Friedkin Group is a privately held family of companies and brands that span the automotive, entertainment, hospitality, sports and investment industries. Ken has decades of experience leading data and analytics teams. Early in his career, he worked with a classic stat software called SPSS. Welcome, Ken.

Ken Elliott:

Thanks, Chad. Good to be here.

Chad Watt:

Ken, tell us about how you got started in data and analytics and why you've stuck with it.

Ken Elliott:

As you mentioned, it was SPSS and I was in grad school using SPSS in my degree and it just sat in Chicago. So it just happened to be the local company. So I took a job there to get myself through grad school, started in tech support. This is when Windows 3.1 came out. So we were teaching a lot of people how to use the mouse and it was just a lot of fun. But the... Got there at the right time. The software company decided to branch out into professional services. So I had the opportunity to build and lead the training department as well as the professional service or consulting department within the organization. And so we started to use statistics and mathematics to help businesses solve business problems. So I finished my degree in industrial psychology, stayed with SPSS for 14 years and wound up becoming the vice president of professional services and solutions and just really fell in love with applying mathematics to solve business problems.

Chad Watt:

Tell me one thing that has changed from that time.

Ken Elliott:

I think a lot has changed. Back then, we called it statistics, and then we started to call it data mining, and then we started to call it predictive analytics, and then it was, what was it, big data, big data analytics, and then machine learning, and now AI. And what's ironic about it is a lot of the core capabilities around data ingestion and analytics and data cleanliness and predictions and the use of mathematics has been pretty constant for the past 30 years. I think, obviously, compute changes, cloud capabilities have changed, applications have changed, but the core math around data, data science, analytics, machine learning, artificial intelligence has been the same. I think the biggest change, quite ironically, has come in the last 12 to 18 months around generative AI. I think that's been really, in my opinion, the tectonic change. Everything else has just kind of been incremental. But the good part is all the basics are still there around how to manage information and turn it into business insights and business value.

Chad Watt:

How are you using data and analytics today at The Friedkin Group.

Ken Elliott:

We're using analytics to do route optimization of our delivery vehicles. We're using AI and analytics to do claims adjudications, we have a financial services arm, so we're automatically adjudicating claims. We are using analytics and optimization algorithms to figure out how do I take a thousand Toyotas a day and distribute them across these five states, 156 dealers to know where to put the right truck or car in the right color with the right options that's gonna sell the fastest at the dealers based on their inventory. So it's this massive optimization problem that historically had been done through spreadsheets and gut and experience. But now we're able to do it at scale with mathematics. And now more recently looking into leveraging agents and AI agents to handle back-office business processes and document management and dozens of other use cases.

Chad Watt:

In order to really put data and analytics to work, what are the prerequisites that you require of your business leaders to give them good insight?

Ken Elliott:

These algorithms learn from the data that you give it and hence the domain of data science. So having good quality data, stored in your data is critically important. Making sure, you know, it's garbage in, garbage out, which was more applicable when I was the chief data officer for Waste Management. We used to say we love garbage data. But those are still the foundations of really good analytics is good data and understanding of data. But in the end, those are solvable problems. The algorithms are certainly solvable. The business problem from an analytic and technology perspective and scale and hosting all the tech is solvable. I think the one thing that really differentiates successful implementations from non-successful implementations are really more around business change management, business adoption, culture. These kinds of issues are the ones that really in the end will separate really successful implementations from non-successful implementations. So to answer your question, I think really a prerequisite is really more on the business side, business change readiness, willingness to change a business process, understanding of their business process, being bold, being able to go after… I mean if you're going to do something meaningful and kind of significant with analytics, it's going to have an impact on your business process and the way that you traditionally did things. And so are you ready to take that journey? I think those… really understanding the business framework within which you're delivering solutions is critical. All the other stuff you can figure out. It's the non-technical stuff that's the hardest.

Chad Watt:

Businesses produce abundant data. What would be the most persistent challenge in putting that data to work?

Ken Elliott:

I'm a huge advocate for the concept of treating data as a corporate asset, literally an asset that adds value to your corporation, just like your vehicle fleet or your real estate or your employees. This is an asset that you have in your company that has value to it. And the more and more we advance the field of analytics and AI and AI agents and these capabilities become more commoditized, really a differentiator for companies is going to be the data that they have, the information that they have, the unique perspective on how they weight certain factors and certain values that represent their strategic advantage. So treating data as an asset means investing in it, means making sure you have good data quality, means that you've got good data definitions, you have stewards within the business that feel an ownership, pride of ownership, if you will, in the data that is within their domain space. So investing in the data, managing it as an asset, stewarding it, monitoring its quality, protecting it, it becomes, to your point, more challenging as we have more and more data. I mean, we've struggled managing structured data for the past 30 years. And now with the advent of generative AI, we now have the responsibility to manage all of that unstructured data that we've been ignoring that's spread out in SharePoint sites within your organization. That potentially represents great value, but also potentially represents great risk that you don't even have awareness of. What kind of content is sitting out there, what information shouldn't be shared that's available that we're now discovering is out there. So I think it's not only the volume of data, but the type of data now that we can reasonably get more value out of with all these generative AI capabilities puts more of a burden on our ability to better manage data to be able to get the value out of it.

Chad Watt:

Now that AI tools have made data more appealing and popular, what are some of the new headwinds that you as a data professional have to deal with?

Ken Elliott:

These kinds of constructs that we are now managing, especially when we start looking at ethical use of information and how an agent is going to communicate with your customers or communicate with your employees, they're going to represent your values. These are data constructs that do represent a new challenge and how do you manage these? How do you manage the voice of your brand? It's one thing to have filters on your data and manage what information goes in and out because it's much more of a discrete set of information that you're managing. But when you get to these more decisions and values and policies and ethics and voice and brand, that's information too that you're going to be feeding and nurturing within these agents that will be running our businesses in the next five years. How do you manage that? And I think that's going to be something we're going to discover together as we go through this venture.

Chad Watt:

So AI changes rapidly, and we keep hearing that this changes everything. Now, in the face of all this rapid change, what are some of the constants of AI and data?

Ken Elliott:

In building these agents and deploying these AI solutions, 80% of what's underneath it is still constant. So as we've talked about already, data is key. So how we manage data, we store it, how we monitor its quality, that's all constant. The software development lifecycle is still the same. You still have to dev and you still have to test and you still have to move into production. So all of these agents, they're mostly software, very little when it comes to data science that are in these agents. Data science is really happening at the LLMs through the source providers, but companies like us that are using those LLMs and putting them into practice, which really is more software engineering around building the applications that these agents complete these workflows within.

There's a massive amount of best practice that we can bring to the table to make sure that these agents are successful. And I think the companies that know how to do software development well and can build enterprise systems well are going to do well with agents.

Chad Watt:

When it comes to data and analytics at The Friedkin Group, how do you measure success?

Ken Elliott:

I measure success as markers along a maturation journey within an organization from kind of experimentation to acceptance to adoption to championship and then ultimately monetization. How do we take this and create new revenue streams or new business opportunities with these capabilities? But, that's kind of at the macro level to answer your question about success. It's like, are we growing this capability and are we really maximizing what's the art of the possible with this capability within an organization? At the minor level, it's anything from how many hours did we save in a business process that we've changed? How much faster are we able to complete a cycle to create better customer experience? There's highly measurable, very traditional kind of metrics that kind of guide our way and really allow you to earn your right to do more. But really at the macro level, I think it's when the conversation switches and the business is really banging on your door, demanding more is when I find it to be successful.

Chad Watt:

Can you tell us about one kind of data project that's kind of at the top of your excitement list right now?

Ken Elliott:

I think one that I found just kind of mathematically fascinating was we have what's called a vehicle processing center. So every day, 1,000 or so cars show up near here in Houston, near the airport. And we have to process these cars and then get them delivered to 156 dealers, as I mentioned before. And this vehicle processing center, there are multiple buildings, whether or not this car needs to have tires added to it, if you have upgraded tires or put the film protection on. So there's like body shop work. Anyway, there's five different buildings that do various things to the car in order to get it into through production and out to the dealers. And there's any number of combinations of things based on what car you ordered and what options you've asked for. So their challenge is how do we get these cars through so that I don't load up all of the big Tundra truck work in building one and then building two is starved for work because they're waiting for the vehicles to come through. How do I slip through the right cars so that I'm maximizing the efficiency and effectiveness of every building and all the technicians across these five buildings so that I can get as many cars through this facility as possible? And you have to factor in available parts. You have to factor in number of technicians and technician skills. You have to factor in the orders of that day, because that's a different, it looks different every day, every hour, you know you've got a different set that you're pulling from is available inventory and getting these vehicles through to the point where they're ready for transportation. So anyway, the complexity, the predictive power within each one of those models and then combining them into an optimization model, which is saying here's how we should pick the next 10, the next 10, the next 10. It was just a really fun mathematical challenge that the team stepped up to and delivered great work.

Chad Watt:

Thank you for your time, Ken.

Ken Elliott:

It's been great. Enjoyed the conversation.

Chad Watt:

The Infosys Knowledge Institute podcast is part of our collaboration with MIT TechReview. Visit our content hub at technologyreview.com to learn more. And be sure to follow the IKI podcast wherever you get your podcasts. You can find more details in our show notes and transcripts at Infosys.com/IKI. That's in our podcast section. Thanks to our producers, Christine Calhoun and Yulia De Bari. Dode Bigley is our audio technician, and I'm Chad Watt with the Infosys Knowledge Institute, signing off. Until next time, keep learning and keep sharing.